Friday, January 21, 2011

The riddle of shear failure

Long ago, when I was an engineering student at Vrije Universiteit Brussel, I did not even realize that my research topic (of which I think I probably could spend a lifetime in researching it) was actually an unanswered question.

I remember taking Concrete Structures in Brussels. This class is the only concrete design class which is offered to civil engineering master students. No offense, I'm only pointing this out to show how different (or less design oriented and much more math and basic principles oriented) the Belgian engineering education is.
When the topic of shear in reinforced concrete beams was discussed, we quickly looked at the equation for the concrete part Vc and then immediately went into two methods of determining the necessary amount of stirrups. The superposition of the concrete part and the steel part was not questioned. The Eurocode 2 formula for the concrete contribution was explained term by term: k is the factor to take into account the size effect,.. To me, it appeared as if there was no problem at all with shear. We have a design formula for shear in beams, which is the holy grail for all shear design.

Two and a half years later, I arrived at Georgia Tech with my two volumes of lecture notes on reinforced concrete from Brussels. Together with my advisor from Georgia Tech, I looked at the material I had covered previously. He looked at the material, and every now and then he would say how "French" my material appeared to him. Even though, according to him, I had covered all topics, he advised me to take the master's course in reinforced concrete, to get used to the strange units and the ACI code.
When shear in concrete was treated, I heard about exotic mechanisms as "aggregate interlock" and "dowel action" for the very first time. I saw an equation for the concrete part Vc which did not look like the Eurocode 2 formula at all. In the lecture notes, the graphs from the ACI committee 326 from the 1960s were shown in which the ACI code formula which was proposed then was compared to a number of shear tests on beams. The scatter was still very large, and gave a coefficient of variation of (order of magnitude) 20%.
For one second the idea crossed my mind that this was because the ACI code formula was much easier and more compact than the Eurocode 2 formula, which looked more exact to me. But shortly afterwards, I started to realize that our current design methods for shear in beams are empirical methods. These methods are the result of shear tests, carried out on small, slender, highly reinforced concrete beams. Extrapolating the results of this types of beams could be questioned. It is therefore not unlogical that it became common practice to use generally conservative rules for shear, to avoid the sudden shear failure and make sure beams (and other structural concrete elements) fail in flexure, since signs of distress appear at load levels below the failure load.

After my job interview at TU Delft, I started to think about the topic for my master's research project course. In correspondence with my future advisors from TU Delft and my advisor from Georgia Tech, we decided to study punching shear in slabs. This problem is related to shear in beams, but works in two dimensions (as a slab has an extra dimensions as compared to a beam). I discovered how much we actually don't know about shear and torsion, and every paper I read just raised more questions. I found it quite exciting to discover that there are still so many questions to be researched.

A year and 5 months ago I started my research at TU Delft. I'm studying both shear in beams and punching shear in slabs and try to see how these mechanisms are interrelated and can occur in bridge decks. Every day I'm learning more, and every day I am formulating more questions to be solved.

However, when I try to explain some of my former classmates from Brussels what I am doing in Delft, I only get some blank stares. Is shear a problem? We have an equation for that in Eurocode 2! And then I tell them about the absurdly high scatter I get when I compare my test data to the calculated values from the code, after which I usually receive a very puzzled look.

* The title of this post is taken after G.N.J. Kani's famous article published in the ACI Journal Proceedings from 1964 (The riddle of shear failure and its solution)


  1. Shear is still a riddle, no doubt. Actually can there be a way of making Finite element procedures to take test results without actually performing tests through computer aided simulations and incorporating those results into empirical relations.
    Well i too am working on minimum shear reinforcement criteria in ACI 318-08. its too much left to discover yet.

  2. Sounds interesting! Which university are you at?

  3. National University of Science and Technology Islamabad, Pakistan. MS Almost over, will be applying for PhD soon. And perhaps i'd switch to the finite element analysis of concrete structures and then compare the FE analysis of similarly modeled beams with those that i had tested. overlaying the Load-deflection curve of tested and modeled results.

  4. the committee dealing with this stuff is ACI-ASCE 447

  5. Great - I'd love to read some of your work once it's out :)
    I'm not using FEA myself, but my experiments are being modeled by our colleagues of the structural mechanics research group, as well as by a student assistant and a MS thesis student (who is actually defending today).

  6. My best wishes for the defence of student.

    if you get time you can read about it on ACI journal paper "Effect of Longitudinal reinforcement and shear span to depth ratio on minimum shear reinforcement ratio" by Lee and Kim, they actually pointed out that by increase of Long. Reinforcement the requirement of min shear reinforcement also increases which aci has not provided provisions yet. And what i am doing here is to prove it via mathematical form (same casting beams, testing them, playing with load deflection diagrams and all that cumbersome procedures). After that next step i am planning to do is to prove it by FEA modeling (though i dont have that strong background in using ansys, nor do people in our country know much about how to model quasibrittle material like concrete in FEA modeling). But for getting enrolled for phd abroad i have to produce a paper atleast.
    and as soon as i complete defence, you'd be the first one to receive the electronic copy of thesis write-up

  7. I've come across that problem previously, but I didn't read much more about it since I am dealing with slabs which typically do not contain shear reinforcement.
    Good luck with writing up your thesis - I am looking forward to reading your results :)